Compal Electronics Inc. - BIOS and OS Tuning Descriptions for SPEC CPU2017 Submissions

Operating System Tuning Parameters

OS Tuning

ulimit:

is a command used to set or check user limits on system resources such as memory, CPU, and the number of open files.Below are common usages of ulimit:

• ulimit -s : Sets the stack size limit.
• ulimit -l : Sets the maximum amount of memory that can be locked into RAM.
• ulimit -n : Limits the maximum number of open file descriptors.
• ulimit -u : Limits the maximum number of user processes that can be created.

irqbalance:

irqbalance is a Linux background service that distributes hardware interrupts across multiple CPU cores to prevent overloading a single core and improve system performance.

Performance Governors (Linux):

are one of Linux's CPU frequency scaling mechanisms, used to determine how the CPU frequency should be managed.Simply put, it controls "how fast the CPU should run under different conditions."Common CPU governors include::

• cpupower frequency-set -g performance : Keeps the CPU running at the highest frequency for maximum performance (but consumes more power).
• cpupower frequency-set -g powersave : Keeps the CPU running at the lowest frequency to save power at the cost of performance.
• cpupower frequency-set -g ondemand : Dynamically switches between high and low frequencies based on system load.
• cpupower frequency-set -g conservative : Similar to ondemand, but changes frequency more gradually for smoother transitions.
• cpupower frequency-set -g schedutil : Adjusts frequency based on information from the task scheduler, balancing performance and power saving.

--governor,-g :

When set to performance, the CPU will always operate at its maximum frequency to deliver the highest computing performance.This will improve overall system performance.

Many companies execute the following command when conducting system performance testing to ensure that the CPU operates at its maximum frequency.

• cpupower frequency-set -g performance

tuned-adm:

is a command-line tool used to manage performance tuning settings on Linux systems. It allows users to select predefined tuning profiles that automatically adjust CPU, power saving, I/O, and network parameters according to the system’s intended usage, optimizing either performance or energy efficiency.The following four are the most commonly used parameters:

• throughput-performance: The throughput-performance profile is optimized for workloads requiring high data throughput. It keeps the CPU at maximum frequency, disables power-saving features, and tunes IRQ and I/O settings to enhance processing efficiency. Ideal for database and storage servers.
• latency-performance: The latency-performance profile targets minimal latency. It disables most power-saving options, keeps the CPU at maximum frequency for long durations, and enhances NUMA locality and IRQ distribution. Best suited for real-time or high-performance computing environments.
• balanced: The balanced profile is a general-purpose setting that provides a compromise between performance and power savings. It dynamically adjusts CPU frequency based on system load and enables moderate power-saving features. It is suitable for general servers or systems without specific performance requirements.
• powersave: The powersave profile prioritizes energy efficiency by setting the CPU to its lowest frequency and enabling all available power-saving features. It is suitable for laptops or idle servers where performance is not critical and power reduction is desired.

Firmware / BIOS / Microcode Settings

Enable LP [Global] (Default = ALL LPs):

Enable LP [Global] to represent the number of logical processors (LP). LP is a term that is common for cloud instances to represent a vCPU.Values for this BIOS option can be:

• ALL LPs
• Single LP

LLC dead line alloc (Default = Enable):

It is a BIOS setting that controls whether a CPU cache line, when it is about to be written back to memory (i.e., a dead-line), is still allowed to be allocated into the LLC (Last Level Cache, typically L3 Cache).Values for this BIOS option can be:

• Disable
• Enable

Energy Efficient Turbo (Default = Enable):

is a technology that allows the processor to automatically boost its frequency based on workload while maintaining energy efficiency.Values for this BIOS option can be:

• Disable
• Enable

Enhanced Halt State (C1E) (Default = Enable):

is a mechanism that allows the processor to automatically reduce voltage and frequency during idle periods to enter a low-power state, while still enabling a quick return to full operation to balance power saving and performance.Values for this BIOS option can be:

• Disable
• Enable

LLC Prefetch (Default = Disable):

The LLC prefetcher is a feature that allows the processor, under a non-inclusive cache architecture, to prefetch data directly into the Last Level Cache (LLC) to improve memory access efficiency, though in some cases, disabling it may actually enhance performance.Values for this BIOS option can be:

• Disable
• Enable

AMP Prefetch (Default = Enable):

The AMP prefetcher is a mechanism that predicts future memory access patterns based on delta sequences between cache accesses and prefetches data into the mid-level cache (MLC), enhancing data retrieval efficiency without issuing LLC prefetches.Values for this BIOS option can be:

• Disable
• Enable

Patrol Scrub (Default = Enable at End of POST):

is a background memory scanning and correction mechanism. It periodically and proactively scans each bit of system memory to detect correctable ECC errors, and attempts to automatically repair them when such errors are found.Values for this BIOS option can be:

• Disable: Completely disables the Patrol Scrub mechanism.
• Enable at End of POST: After completing the POST (Power-On Self Test), the system performs a full memory scan.

Homeless Prefetch (Default = Auto):

is a mechanism that allows data to be prefetched into the L2 cache when cache resources are limited to reduce latency, but it may impose a burden on uncore resources under certain workloads.Values for this BIOS option can be:

• Auto
• Disable
• Enable

DCU Streamer Prefetcher (Default = Enable):

is an L1 data cache prefetcher. Recommended default setting is Enabled. In some cases, setting this option to disabled can improve performance. Values for this BIOS option can be:

• Disable
• Enable

Package C State (Default = Auto):

The system automatically manages C-state transitions based on workload and platform policies.Values for this BIOS option can be:

• C0/C1 state:Only allows the processor to enter basic idle states (C1).Provides fast wake-up latency with minimal power savings.Recommended for latency-sensitive or real-time workloads.
• C2 state:Enables a deeper idle state than C1, offering more power savings at the cost of slightly higher latency.
• C6(non Retention) state:Allows the processor to enter a deep power-saving state where core power and cache are flushed.“Non Retention” means cache contents are not preserved, leading to higher wake-up latency.
• No Limit:The processor is allowed to enter the deepest supported C-state available on the platform.Maximizes power savings, but may increase wake latency.
• Auto: Automatically manages Package C-states based on workload.

ENERGY_PERF_BIAS_CFG mode (Default = Balanced Performance):

It is a hint provided by the processor to the BIOS or operating system to guide the trade-off between power efficiency and performance.Values for this BIOS option can be:

• Balanced Performance: Prioritizes performance while moderately considering power efficiency; a compromise between full performance and power saving.
• Performance: Force performance priority; the CPU maintains higher frequencies, disregarding power-saving considerations.
• Balanced Power: Prioritizes power saving while maintaining a reasonable level of performance; suitable for reducing power consumption without heavily sacrificing performance.
• Power: Prioritizes power saving; the system aggressively reduces CPU power consumption, even at the expense of performance.

Virtual Numa (Default = Disable):

Divide physical NUMA nodes into evenly sized virtual NUMA nodes in ACPI table. This may improve Windows performance on CPUs with more than 64 logical processors.Values for this BIOS option can be:

• Enable
• Disable

Hardware P-States (Default = Native Mode):

(HWP) are the key components of HWPM which deals with P-state control. HWP works by freeing the operating system from making direct frequency decisions.Values for this BIOS option can be:

• Disable: fix the core frequency to P1.
• Native Mode:the OS provides high level hints or constraints to hardware and hardware makes pstate decisions by observing core statistics and OS-provided guidance or hints.
• Out Of Band Mode:hardware autonomously makes pstate decisions with OS completely excluded from this process.
• Native Mode With No Legacy Support:Completely disables the legacy P-State interface, allowing the processor to use HWP exclusively to manage core frequency and power.

Hardware Prefetcher(Default=Enable):

Hardware Prefetcher: (a.k.a. MLC Streamer Prefetcher) is an L2 cache prefetcher. Recommended default setting is Enabled. In some cases, setting this option to disabled can improve performance. Values for this BIOS option can be:

• Disable
• Enable

Adjacent Cache Prefetch(Default=Enable):

Adjacent Cache Prefetch: (a.k.a. MLC Spatial Prefetcher) is an L2 cache prefetcher. Recommended default setting is Enabled. In some cases, setting this option to disabled can improve performance. Values for this BIOS option can be:

• Disable
• Enable

DCU IP Prefetcher(Default=Enable):

DCU IP Prefetcher: the DCU Instruction Pointer (IP) prefetcher is an L1 cache prefetcher. Recommended default setting is Enabled. In some cases, setting this option to disabled can improve performance. Values for this BIOS option can be:

• Disable
• Enable

Turbo Mode(Default=Enable):

Turbo Boost Technology requires a PC with a processor with Intel Turbo Boost Technology capability. Intel Turbo Boost Technology performance varies depending on hardware, software, and overall system configuration.Values for this BIOS option can be:

• Disable
• Enable

Power Performance Tuning(Default=OS Controls EPB):

Power Performance Tuning is a BIOS setting that determines whether the processor’s performance and power efficiency preferences are managed by the operating system, BIOS, or an external controller.Values for this BIOS option can be:

• OS Controls EPB : Power and performance tuning are managed by the operating system.
• BIOS Controls EPB:Power and performance tuning are managed by the BIOS settings, ignoring OS-level adjustments.
• PECI Controls EPB:Adjusted through PECI (Platform Environment Control Interface) by external controllers such as the BMC.

Performance Mode(Default=Balanced):

It is a common setting in server BIOS used to control the system’s overall performance optimization strategy and power behavior.Values for this BIOS option can be:

• Balance : :is the default setting for most users, providing a compromise between performance and power efficiency.
• Performance:allows your system to operate at maximum performance, but power consumption and temperature will increase.
• Custom:is intended for advanced users or server administrators who need to manually fine-tune detailed parameters.

Latency Optimized Mode(Default=Disable):

The highest performance mode (default on previous generations) where core and uncore frequencies are running up to their maximum limits within the RAPL budget. This mode is not performance-per-Watt optimized across the load line. Values for this BIOS option can be:

• Disable
• Enable

NUMA(Default=Enable):

Each CPU (or each group of CPUs) has its own "local memory," which can be accessed faster. Accessing "remote memory" across CPUs is slower. When NUMA is enabled, the operating system can arrange thread execution and memory allocation based on the "physical locality" of CPUs and memory, reducing latency and improving performance. Values for this BIOS option can be:

• Disable
• Enable

Opportunistic LLC to SF Migration(Default=Disable):

When data is present in the LLC (Last Level Cache, such as L3 Cache), the system may opportunistically migrate it to the Snoop Filter (SF) to optimize and accelerate inter-core data sharing.Values for this BIOS option can be:

• Disable
• Enabled

Last updated May 7, 2025.